JPH06259534A - Person identifying device - Google Patents

Abstract

PURPOSE:To provide a person identifying device capable of preventing the accuracy of identification from being reduced due to the influence of head posture, face expression, etc., and improving the measuring accuracy of a featured value for judging an effective frame. CONSTITUTION:The person identifying device is provided with a means 1 for inputting picture data, a means 2 for detecting a face area from the picture data, a means 3 for detecting a featured value from the face area, a means 4 for judging whether the featured value is effective or not by comparing it with a reference value, a means 6 for storing featured values judged as effective ones for plural frames, a means 7 for obtaining an identifying featured value by executing the operation of featured values for plural frames, and a means 8 for identifying a person by comparing the identifying featured value with a previously stored face featured value.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a person authentication device for extracting a feature amount from a face image of a person and performing authentication.

[0002]

2. Description of the Related Art A conventional person authentication device is characterized in that a plurality of image data are used to detect a feature amount from image data suitable for authentication in order to improve authentication accuracy. FIG. 11 shows a block diagram of a conventional device. Here, the feature amount detection unit 3 detects a predetermined feature amount, and the valid frame determination unit 4 compares the value with each value of the determination reference value recording unit 10 to obtain an image satisfying the predetermined condition. Make a selection. And
When the image satisfies a predetermined condition, the person authentication device detects the characteristic amount for authentication and authenticates the person. When the image does not satisfy the predetermined condition, the person authentication device performs the validity determination of the next image data.

The conditions for image selection are photographing conditions, head posture,
It is whether or not the facial expression is quantitatively measured and the value is determined or the calculation result using the value satisfies the determined condition.

For example, a front face is defined with respect to the posture of the head, and the validity of the image data is determined by performing an operation on a predetermined characteristic amount. The posture of the head was decomposed into three movements shown in FIG. With respect to the A axis in FIG. 12, the rotation can be detected by obtaining and comparing the distances D1 and D2 from the center of the mouth to both cheeks shown in FIG. The B axis can be detected by comparing the height positions of the centers of the left and right eyes shown in FIG. But,
The movement in the C-axis direction cannot be determined from one piece of image data, and the movement in this direction affects the authentication accuracy.

Alternatively, the state of opening and closing the eyes and opening and closing the mouth is
The maximum value and the minimum value were set for each value of the feature amount related to the eyes and mouth, and the effective image data was judged depending on whether or not it was within the range. However, because the effective range is wide, the accuracy of judgment was not sufficient, which affected the authentication. As described above, in the conventional apparatus, there is a state in which valid image data cannot be accurately detected, and there is a drawback in that the authentication accuracy decreases.

[0006]

As described above, in the conventional apparatus, it is difficult to eliminate the influence of the head posture and facial expression from the characteristic amount of one piece of image data determined to be effective, and the authentication is performed. There was a drawback that the accuracy was lowered.

An object of the present invention is to provide a person authentication device which eliminates the deterioration of the authentication accuracy caused by the influence of the posture of the head and facial expressions and further improves the measurement accuracy of the feature quantity for determining the effective frame.

[0008]

According to the present invention, means for inputting image data for performing person authentication in frame units, means for detecting a face area from the image data, and detection of a feature amount from the face area. Means, a means for determining whether or not the feature quantity is valid by comparing the feature quantity with a reference value, a means for storing the feature quantity determined to be valid for a plurality of frames, and a feature for the plurality of frames A means for obtaining an authentication characteristic amount by performing a calculation on the amount, and a means for authenticating the person by comparing the authentication characteristic amount with a face characteristic amount stored in advance. To do.

Further, according to the present invention, means for inputting image data for performing person authentication in frame units, means for detecting a face area from the image data, means for detecting a feature amount from the face area, and A means for determining whether or not the feature quantity is valid by comparing the feature quantity with a reference value, a means for storing the feature quantity determined to be valid for a plurality of frames, and a calculation for the feature quantity for the plurality of frames. A unit that changes the reference value by rubbing, a unit that determines whether the feature amount is valid by comparing the feature amount detected from the face area with the changed reference value, and that the feature amount is determined to be valid. And a means for authenticating the person by comparing the feature amount with a face feature amount stored in advance.

[0010]

According to the present invention, the characteristic amount of the image data obtained by the effective frame determination is recorded for a predetermined number of sheets (a plurality of frames), the value is calculated, and the obtained result is obtained. It is used for authentication.

Alternatively, the reference value for valid frame determination is changed using the obtained calculation result, and the image data satisfying the new condition is selected to improve the authentication accuracy.

Alternatively, the characteristic amount of the image data selected by the new reference value and the determined number of sheets are recorded, and a predetermined arithmetic operation is performed to obtain the characteristic amount for authentication, thereby improving the authentication accuracy. ing. By performing such processing, it is possible to provide a device in which the authentication accuracy does not decrease due to changes in the posture of the head and changes in facial expressions, and it is possible to expect improvement in the detection accuracy of the feature amount.

[0013]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing the configuration of the first exemplary embodiment of the present invention. The image data input unit 1 converts a signal input from a camera into a digital signal and converts it into a predetermined signal format (luminance signal, color signal, etc.).
The face area detection unit 2 detects a face area from the image data based on the color signal.

The feature amount detecting section 3 examines the face area and detects the position of the feature point of the face shown in FIG. 2 by using the edge information.
The position information, size, and distance of the characteristic points shown in FIG. 3 are obtained and sent to the effective frame determination unit 4.

The effective frame determination unit 4 performs a predetermined calculation using the data from the feature amount detection unit 3 and compares the result with the value of the determination reference value recording unit 10,
It is determined whether the input image is an image satisfying a predetermined condition based on the result of data comparison, and a determination result signal is sent to the control unit 5.

If the determination result is invalid, the control unit 5 sends an input instruction signal to the image input unit 1, and if the determination result is valid, the number of valid image data recorded by the control unit 5 is increased and the effective feature is determined. A feature quantity recording instruction signal is sent to the quantity recording section 6, and an input instruction signal is sent to the image input section 1. When the number of valid image data has reached a predetermined number, a calculation start instruction signal is sent to the authentication feature amount calculation unit 7. The control unit 5 is composed of a CPU, a ROM, a RAM, etc., and generates a signal for operating this device by following a predetermined processing procedure. Effective feature amount recording unit 6
Is a means for holding the characteristic data for a plurality of predetermined frames.

In the authentication feature quantity computing unit 7, the effective feature quantity (the size of each feature,
Using the data (such as the positional relationship), an authentication feature vector is obtained based on a predetermined calculation, and the result is sent to the authentication unit 8.

As an example of the calculation, if the average value of the determined number of sheets of each value is calculated and the result is used as the authentication vector, the influence of the measurement error can be suppressed. Alternatively, a system is conceivable in which image data is extracted from the face area using the obtained value, the value is converted on the frequency axis by DFFT, etc., and the value is used as the authentication vector.

Here, as an example of the calculation for suppressing the influence of the facial expression, FIG. 4 showing the change of the eye size, the change of the eye size recorded in the effective feature amount recording section 6 as the effective feature amount is shown. This will be described with reference to FIG. 5 which is shown and FIG. 6 which shows a histogram of the effective feature amount.

FIG. 4 is a diagram showing how the size of eyes changes with time when blinking is performed. The vertical axis shows the size of eyes and the horizontal axis shows the change with time. The value H1 is the true size of the eyes, and H2 and H3 are the reference values for judgment.
From the time t1 to the time t2, since the judgment reference value is not satisfied, it is not recorded in the effective feature amount recording unit 6 as an effective feature amount, and other values are effective values and are recorded as effective value. Recorded in 6.

FIG. 5 is a diagram showing the values recorded in the effective characteristic amount recording section 6 as the effective characteristic amounts, in which the vertical axis represents the size of the eyes and the horizontal axis represents the change with time. Comparing the recorded value with the true value H1, many points with large errors are included. The point A is a measurement error, which is larger than the original value. A histogram of eye size is created based on this data.

FIG. 6 is a histogram of the feature quantity recorded in the effective feature quantity recording unit 6, in which the vertical axis represents frequency and the horizontal axis represents the eye size value. The frequency of this histogram is examined from the larger one, and the value having a predetermined size is defined as the size of the eye. Alternatively, a measurement value with a small measurement error is obtained by obtaining the average value of only the data satisfying the values before and after this value.

The authentication accuracy can be expected to improve by performing the above-described calculation in the authentication feature amount calculator 7. Also,
In the case described below, the above calculation is performed by the reference value calculation unit 11
And sends the result to the valid frame determination unit 4.

The authentication unit 8 calculates the authentication vector and the reference value recorded in the face feature amount recording unit 9, and if the calculation result is less than a predetermined value, authentication is performed, and the control unit 5
The authentication result signal is sent to and the result is output.

When the calculation result is equal to or more than the determined value, the authentication invalid signal is sent to the control unit 5. The control unit 5 resets the value of the number of effective images to 0, discards the data of the effective feature amount recording unit 6, and sends the image input unit 1 an input instruction signal for obtaining new image data.

As described above, according to the first embodiment, a plurality of frames of effective feature quantities regarding the size of the eyes and mouth are detected from the face area, and the average value of these feature quantities is used as the authentication feature quantity. Therefore, the accuracy of the authentication rate can be improved as compared with the conventional device. FIG. 7 is a block diagram showing the configuration of the second embodiment of the present invention. Here, the points different from the first embodiment will be described according to the processing procedure.

The reference value calculation unit 11 calculates the effective feature amount for a plurality of frames recorded in the effective feature amount recording unit 6, and sends the result to the effective frame determination unit 4. The control unit 5 sends a determination reference value change signal to the valid frame determination unit 4. The valid frame determination unit 4 changes the reference value for valid frame determination according to a predetermined means. The control unit 5 sets the number of effective frames to 0 and sends an input instruction signal to the image data input unit 1.

Thereafter, the valid frame determination is performed using the new reference value, not the previous reference value recorded in the determination reference value recording section 10, and the image data satisfying this value is selected to select the selected image. Authentication is performed by obtaining an authentication vector from only the feature amount of data. Here, an algorithm for detecting the rotation about the C-axis in FIG. 12 which is a problem in the conventional example will be described with reference to FIG.

The image input device inputs image data of a state in which a three-dimensional object is projected in two dimensions, and the measured value in the height direction is the length of the object projected in two dimensions. I am measuring. Therefore, assuming that the actual length is L, the apparent value L ′ to be measured can be expressed by the following equation, where θ is the angle at which the object is in the vertical direction. L '= L COS θ

From this equation, when the apparent length is maximum, the head is perpendicular to the ground (Z = 0). Therefore, for the rotation of the C axis, if the maximum value is detected from the value of the nose length shown in FIG.

FIG. 9 is a diagram showing how the nose length data of the person to be authenticated changes. Here, the person is
It is recorded that the data corresponding to the length of the nose changes momentarily by rotating the head in the C-axis direction. Of these, the maximum value is regarded as the length of the nose when this person faces the front accurately, and only this maximum value is detected. Then, the reference value (related to the general nose length) is changed with this maximum value as the true nose length when determining the effective frame of the person. Therefore, after that, the image data only when the person faces the front is selected as the effective frame from the new reference value.

That is, if this value is set as a new reference value of the valid frame determination section 4 and image data satisfying this condition is selected, image data not rotated about the C axis can be obtained. By using the characteristic amount of the image data at this time as an authentication vector and performing authentication by the authentication unit 8, it is possible to provide an apparatus for performing authentication without being affected by rotation in the C-axis direction.
FIG. 10 is a block diagram showing the configuration of the third exemplary embodiment of the present invention. Differences from the first and second embodiments will be described according to the processing procedure. Similar to the second embodiment, the reference value calculation unit 1
In 1, the effective feature amount is calculated to obtain a new reference value for determination. The valid frame determination unit 4 selects an image frame that satisfies the new reference value.

The control unit 5 receives from the valid frame determination unit 4
When the signal of the validity determination is received, a recording instruction signal is sent to the effective feature amount recording unit 6 to increase the number of effective frames by one,
An input instruction signal is sent to the image data input unit 1. When the number of valid frames reaches the specified number, a calculation start signal is sent to the authentication feature amount calculation unit 7.

In the authentication characteristic amount calculation 7, an average value is calculated from the respective characteristic amounts of the effective frames that have reached the specified number, and this is used as the authentication characteristic amount. Then, the authentication unit 8 obtains the authentication result by comparing with the face feature amount stored in advance.

That is, the third embodiment is a technique for obtaining the authentication feature amount from the effective feature amounts for a plurality of frames in the first embodiment and the effective feature amount for a plurality of frames in the second embodiment. It is a combination of technologies for changing the reference value from. Therefore, with this device, you can select image data with good eye and mouth conditions without rotating about the C-axis.
In addition, since an accurate feature amount can be obtained, it is possible to provide an authentication device having an improved authentication rate as compared with a conventional device.

[0036]

According to the person authenticating apparatus of the present invention, the feature amount can be detected with high accuracy even when the posture or facial expression of the user is changed. Therefore, in the future security system or the like, the effect that the authentication rate for each user is improved can be obtained.

[Brief description of drawings]

FIG. 1 is a diagram showing an overall configuration of a first embodiment of the present invention.

FIG. 2 is a diagram showing facial feature points.

FIG. 3 is a diagram showing a feature amount of a face.

FIG. 4 is a diagram showing how the size of eyes changes with time.

FIG. 5 is a diagram showing how the size of eyes as an effective feature amount changes with time.

FIG. 6 is a diagram showing a histogram with respect to changes in eye size over time.

FIG. 7 is a diagram showing an overall configuration of a second embodiment of the present invention.

FIG. 8 is a diagram showing parameters for obtaining rotation of the head in the C-axis direction.

FIG. 9 is a diagram showing how the nose length changes with the rotation of the head in the C-axis direction.

FIG. 10 is a diagram showing the overall configuration of a third embodiment of the present invention.

FIG. 11 is a diagram showing an overall configuration of a conventional example.

FIG. 12 is a diagram showing the rotation of the head in three directions.

FIG. 13 is a diagram showing parameters for obtaining rotation of the head in the A-axis and B-axis directions.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Image data input unit 2 ... Face area detection unit 3 ... Feature amount detection unit 4 ... Effective frame determination unit 5 ... Control unit 6 ... Effective feature amount recording unit 7 ... Authentication feature amount operation unit 8 ... Authentication unit 9 ... Face Feature amount recording unit 10 ... Judgment reference value recording unit 11 ... Reference value calculation unit

Claims (3)

[Claims] 1. A unit for inputting image data for performing person authentication in frame units, a unit for detecting a face area from the image data, a unit for detecting a feature amount from the face region, and a unit for calculating the feature amount. A means for determining whether or not the feature quantity is valid by comparing with a reference value; a means for storing the feature quantity determined to be valid for a plurality of frames; and a calculation for the feature quantity for the plurality of frames. A person authentication apparatus comprising: a means for obtaining an authentication characteristic amount; and a means for authenticating the person by comparing the authentication characteristic amount with a face feature amount stored in advance. 2. A means for inputting image data for performing person authentication in frame units, a means for detecting a face area from the image data, a means for detecting a feature amount from the face area, and the feature amount. A means for determining whether or not the feature quantity is valid by comparing with a reference value; a means for storing the feature quantity determined to be valid for a plurality of frames; and a calculation for the feature quantity for the plurality of frames. A means for changing the reference value, a means for determining whether or not the feature amount is valid by comparing the feature amount detected from the face area with the changed reference value, and a feature amount determined to be valid And a means for authenticating the person by comparing with a face feature amount stored in advance. 3. A means for accumulating a plurality of frames of feature quantities determined to be valid by comparing with the changed reference value and performing an arithmetic operation on these feature quantities to obtain an authentication feature quantity. 3. The person authentication device according to claim 2, wherein the means for performing authentication compares the authentication feature amount with a face feature amount stored in advance.